Improved Automatic Edge Detection Scanning for our Camera Scanner, ScannerLens

AtlasV Singapore
3 min readMar 10, 2021

There are undoubtedly many camera scanner applications in the market, from the likes of CamScanner, ScanMaster and Microsoft Office Lens. The concept is simple yet effective — making the mobile phone a portable and competent scanner with automatic edge detection, to replace the old, chunky and user unfriendly old-school scanners. In just a tap, you can scan almost anything you want, as long as you have decent lighting and a flat surface.

One key feature of most camera scanner apps is the automatic edge detection function, which, as the name suggests, detects the edges of your document and marks the scanned area. As this feature is automated, it takes machine learning and research to refine and enhance it.

Our latest version with an improved automatic edge detection

Our Product and Tech Team has been working hard to improve the automatic edge detection for ScannerLens, and the result is a more accurate camera scanning experience for users! Our automatic edge detection works even under shaky or wobbly conditions. As pioneers in the field, we’d love to share our thought-process in this refinement, and how we tested and improved this feature in our latest version update.

Creating a 8-point Grid System

While not entirely related, our Product and Tech Team took learnings from the 8-pt Grid System, one of the most commonly used and enduring Layout Grid Systems out there. You can read more about 8-point Grid System and how it guides design and layout here. We used an even number like 8 to space and size the document, so that the scaling and scanning process becomes easier and more consistent over time through machine learning.

Multiple Scans for Automatic Edge Detection Machine Learning

Harnessing the power of Artificial Intelligence (AI), we conducted machine learning into the automation of edge detection in documents to achieve this refinement. With each scan, we reduced the variance and standard deviation from the optimal, ideal scan, bit by bit. This was done over 100+ times per document scan, and we experimented with different documents of different dimensions and paper material to achieve better results against difference in glare, reflection and dimension of paper materials.

After several tests, our automatic edge detection feature is now optimised to accurately crop to the document’s edges in frame by at least 85% in accuracy. If the automatic edge detection is not accurate enough in your scan, you can always make use of the 8 points to manually adjust and achieve the edge cropping!

We are always working towards making our products better. We welcome your reviews on the App Store. If you have any problems or suggestions for our apps, feel free to contact us here, or check out our detailed Frequently Asked Questions.

Check out ScannerLens: Facebook | Instagram | Twitter | App Store | Website

Originally published at https://scannerlens.com on March 10, 2021.

--

--

AtlasV Singapore

Productivity Apps company based in Singapore. We build innovative apps that change the way we work for the better.